In vitro screening of silver nanoparticles and ionic silver using neural networks yields differential effects on spontaneous activity and pharmacological responses.

Silver nanoparticles (AgNPs) are used in a wide range of consumer and medical products because of their antimicrobial and antifungal properties, and can translocate to the brain following exposure. Therefore, to screen AgNPs for potential impacts on human health, it is essential to examine neural function. The present study examined AgNPs (3 citrate coated, 3 PVP coated; 10-75nm) and AgNO3 effects on spontaneous and pharmacologically-induced neural network function in rat primary cortical cells on multi-well microelectrode array (mwMEA) plates. Baseline activity (1h) was recorded prior to exposure to non-cytotoxic concentrations of AgNPs and AgNO3 (0.08-0.63 and 0.08-1.7µg/ml, respectively). Changes in number of total extracellularly-recorded action potential spikes (total spikes; TS) and active electrodes (AE), relative to controls, were assessed 1, 24, and 48h after exposure to AgNP suspensions or AgNO3. After the 48h recording, the response to a pharmacological challenge with the GABAA antagonist, bicuculline (BIC), was assessed. Only two particles altered neural network function. Citrate coated 10nm AgNP caused concentration-related increases in AEs at 24h. After BIC treatment, PVP coated 75nm AgNP caused concentration-dependent increases in AE. AgNO3 effects differed from AgNPs, causing a concentration-related decrease in AEs at 24 and 48h, and a concentration-related decrease in TS following BIC challenge. Importantly, the direction of AgNO3 effects on neural activity was opposite those of 10nm Ag citrate at concentrations up to 0.63µg/ml, and different from 75nm Ag PVP, indicating ionic silver does not mediate these effects. These results demonstrate that non-cytotoxic concentrations of 10nm citrate- and 75nm PVP-coated Ag NPs alter neural network function in vitro, and should be considered for additional neurotoxicity hazard characterization.

In vitro screening of metal oxide nanoparticles for effects on neural function using cortical networks on microelectrode arrays.

Nanoparticles (NPs) may translocate to the brain following inhalation or oral exposures, yet higher throughput methods to screen NPs for potential neurotoxicity are lacking. The present study examined effects of 5 CeO2 (5- 1288 nm), and 4 TiO2 (6-142 nm) NPs and microparticles (MP) on network function in primary cultures of rat cortex on 12 well microelectrode array (MEA) plates. Particles were without cytotoxicity at concentrations ≤50 µg/ml. After recording 1 h of baseline activity prior to particle (3-50 µg/ml) exposure, changes in the total number of spikes (TS) and # of active electrodes (#AEs) were assessed 1, 24, and 48 h later. Following the 48 h recording, the response to a challenge with the GABAA antagonist bicuculline (BIC; 25 µM) was assessed. In all, particles effects were subtle, but 69 nm CeO2 and 25 nm TiO2 NPs caused concentration-related decreases in TS following 1 h exposure. At 48 h, 5 and 69 nm CeO2 and 25 and 31 nm TiO2 decreased #AE, while the two MPs increased #AEs. Following BIC, only 31 nm TiO2 produced concentration-related decreases in #AEs, while 1288 nm CeO2 caused concentration-related increases in both TS and #AE. The results indicate that some metal oxide particles cause subtle concentration-related changes in spontaneous and/or GABAA receptor-mediated neuronal activity in vitro at times when cytotoxicity is absent, and that MEAs can be used to screen and prioritize nanoparticles for neurotoxicity hazard.

Reconstructing exposures from biomarkers using exposure-pharmacokinetic modeling--A case study with carbaryl.

Sources of uncertainty involved in exposure reconstruction for short half-life chemicals were characterized using computational models that link external exposures to biomarkers. Using carbaryl as an example, an exposure model, the Cumulative and Aggregate Risk Evaluation System (CARES), was used to generate time-concentration profiles for 500 virtual individuals exposed to carbaryl. These exposure profiles were used as inputs into a physiologically based pharmacokinetic (PBPK) model to predict urinary biomarker concentrations. These matching dietary intake levels and biomarker concentrations were used to (1) compare three reverse dosimetry approaches based on their ability to predict the central tendency of the intake dose distribution; and (2) identify parameters necessary for a more accurate exposure reconstruction. This study illustrates the trade-offs between using non-iterative reverse dosimetry methods that are fast, less precise and iterative methods that are slow, more precise. This study also intimates the necessity of including urine flow rate and elapsed time between last dose and urine sampling as part of the biomarker sampling collection for better interpretation of urinary biomarker data of short biological half-life chemicals. Resolution of these critical data gaps can allow exposure reconstruction methods to better predict population-level intake doses from large biomonitoring studies.

Systematic proteomic approach to characterize the impacts of chemical interactions on protein and cytotoxicity responses to metal mixture exposures.

Chemical interactions have posed a big challenge in toxicity characterization and human health risk assessment of environmental mixtures. To characterize the impacts of chemical interactions on protein and cytotoxicity responses to environmental mixtures, we established a systems biology approach integrating proteomics, bioinformatics, statistics, and computational toxicology to measure expression or phosphorylation levels of 21 critical toxicity pathway regulators and 445 downstream proteins in human BEAS-2B cells treated with 4 concentrations of nickel, 2 concentrations each of cadmium and chromium, as well as 12 defined binary and 8 defined ternary mixtures of these metals in vitro. Multivariate statistical analysis and mathematical modeling of the metal-mediated proteomic response patterns showed a high correlation between changes in protein expression or phosphorylation and cellular toxic responses to both individual metals and metal mixtures. Of the identified correlated proteins, only a small set of proteins including HIF-1α is likely to be responsible for selective cytotoxic responses to different metals and metals mixtures. Furthermore, support vector machine learning was utilized to computationally predict protein responses to uncharacterized metal mixtures using experimentally generated protein response profiles corresponding to known metal mixtures. This study provides a novel proteomic approach for characterization and prediction of toxicities of metal and other chemical mixtures.

Use of novel inhalation kinetic studies to refine physiologically-based pharmacokinetic models for ethanol in non-pregnant and pregnant rats.

Ethanol (EtOH) exposure induces a variety of concentration-dependent neurological and developmental effects in the rat. Physiologically-based pharmacokinetic (PBPK) models have been used to predict the inhalation exposure concentrations necessary to produce blood EtOH concentrations (BEC) in the range associated with these effects. Previous laboratory reports often lacked sufficient detail to adequately simulate reported exposure scenarios associated with BECs in this range, or lacked data on the time-course of EtOH in target tissues (e.g. brain, liver, eye, fetus). To address these data gaps, inhalation studies were performed at 5000, 10 000, and 21 000 ppm (6 h/d) in non-pregnant female Long-Evans (LE) rats and at 21 000 ppm (6.33 h/d) for 12 d of gestation in pregnant LE rats to evaluate our previously published PBPK models at toxicologically-relevant blood and tissue concentrations. Additionally, nose-only and whole-body plethysmography studies were conducted to refine model descriptions of respiration and uptake within the respiratory tract. The resulting time-course and plethysmography data from these in vivo studies were compared to simulations from our previously published models, after which the models were recalibrated to improve descriptions of tissue dosimetry by accounting for dose-dependencies in pharmacokinetic behavior. Simulations using the recalibrated models reproduced these data from non-pregnant, pregnant, and fetal rats to within a factor of 2 or better across datasets, resulting in a suite of model structures suitable for simulation of a broad range of EtOH exposure scenarios.

Multi-well microelectrode array recordings detect neuroactivity of ToxCast compounds.

Spontaneous activity in neuronal cultures on microelectrode arrays (MEAs) is sensitive to effects of drugs, chemicals, and particles. Multi-well MEA (mwMEA) systems have increased throughput of MEAs, enabling their use for chemical screening. The present experiments examined a subset of EPA's ToxCast compounds for effects on spontaneous neuronal activity in primary cortical cultures using 48-well MEA plates. A first cohort of 68 compounds was selected from the ToxCast Phase I and II libraries; 37 were positive in one or more of 20 individual ToxCast Novascreen assays related to ion channels (NVS_IC), with the remainder selected based on known neuroactivity. A second cohort of 25 compounds was then tested with 20 originating from the ToxCast Phase I and II libraries (not hits in NVS_IC assays) and 5 known negatives from commercial vendors. Baseline activity (1h) was recorded prior to exposing the networks to compounds for 1h, and the weighted mean firing rate (wMFR) was determined in the absence and presence of each compound. Compounds that altered activity by greater than the weighted change of DMSO-treated wells plus 2SD were considered "hits". Of the first set of 68 compounds, 54 altered wMFR by more than the threshold, while in the second set, 13/25 compounds were hits. MEAs detected 30 of 37 (81.1%) compounds that were hits in NVS_IC assays, as well as detected known neurotoxicants that were negative in NVS_IC assays, primarily pyrethroids and GABAA receptor antagonists. Conversely, wMFR of cortical neuronal networks on MEAs was insensitive to nicotinic compounds, as only one neonicotinoid was detected by MEAs; this accounts for the bulk of non-concordant compounds between MEA and NVS_IC assays. These data demonstrate that mwMEAs can be used to screen chemicals efficiently for potential neurotoxicity, and that the results are concordant with predictions from ToxCast NVS_IC assays for interactions with ion channels.

Evaluation of microelectrode array data using Bayesian modeling as an approach to screening and prioritization for neurotoxicity testing.

The need to assess large numbers of chemicals for their potential toxicities has resulted in increased emphasis on medium- and high-throughput in vitro screening approaches. For such approaches to be useful, efficient and reliable data analysis and hit detection methods are also required. Assessment of chemical effects on neuronal network activity using microelectrode arrays (MEAs) has been proposed as a screening tool for neurotoxicity. The current study examined a Bayesian data analysis approach for assessing effects of a 30 chemical training set on activity of primary cortical neurons grown in multi-well MEA plates. Each well of the MEA plate contained 64 microelectrodes and the data set contains the number of electrical spikes registered by each electrode over the course of each experiment. A Bayesian data analysis approach was developed and then applied to several different parsings of the data set to produce probability determinations for hit selection and ranking. This methodology results in an approach that is approximately 74% sensitive in detecting chemicals in the training set known to alter neuronal function (23 expected positives) while being 100% specific in detecting chemicals expected to have no effect (7 expected negatives). Additionally, this manuscript demonstrates that the Bayesian approach may be combined with a previously published weighted mean firing rate approach in order to produce a more robust hit detection method. In particular, when combined with the weighted mean firing rate approach, the joint analysis produces a sensitivity of approximately 96% and a specificity of 100%. These results demonstrate the utility of a novel approach to analysis of MEA data and support the use of neuronal networks grown on MEAs as a for neurotoxicity screening approach.

Development of multi-route physiologically-based pharmacokinetic models for ethanol in the adult, pregnant, and neonatal rat.

Biofuel blends of 10% ethanol (EtOH) and gasoline are common in the USA, and higher EtOH concentrations are being considered (15-85%). Currently, no physiologically-based pharmacokinetic (PBPK) models are available to describe the kinetics of EtOH-based biofuels. PBPK models were developed to describe life-stage differences in the kinetics of EtOH alone in adult, pregnant, and neonatal rats for inhalation, oral, and intravenous routes of exposure, using data available in the open literature. Whereas ample data exist from gavage and intravenous routes of exposure, kinetic data from inhalation exposures are limited, particularly at concentrations producing blood and target tissue concentrations associated with developmental neurotoxicity. Compared to available data, the three models reported in this paper accurately predicted the kinetics of EtOH, including the absorption, peak concentration, and clearance across multiple datasets. In general, model predictions for adult and pregnant animals matched inhalation and intravenous datasets better than gavage data. The adult model was initially better able to predict the time-course of blood concentrations than was the neonatal model. However, after accounting for age-related changes in gastric uptake using the calibrated neonate model, simulations consistently reproduced the early kinetic behavior in blood. This work provides comprehensive multi-route life-stage models of EtOH pharmacokinetics and represents a first step in development of models for use with gasoline-EtOH blends, with additional potential applicability in investigation of the pharmacokinetics of EtOH abuse, addiction, and toxicity.

Evaluation of multi-well microelectrode arrays for neurotoxicity screening using a chemical training set.

Microelectrode array (MEA) approaches have been proposed as a tool for detecting functional changes in electrically excitable cells, including neurons, exposed to drugs, chemicals or particles. However, conventional single well-MEA systems lack the throughput necessary for screening large numbers of uncharacterized compounds. Recently, multi-well MEA (mwMEA) formats have become available to address the need for increased throughput. The current experiments examined the effects of a training set of 30 chemicals on spontaneous activity in networks of cortical neurons grown on mwMEA plates. Each plate contained 12 wells with 64 microelectrodes/well, for a total of 768 channels. Of the 30 chemicals evaluated, 23 were known to alter neuronal function in vivo ("positives"), including 6 GABAergic and 3 glutamatergic antagonists/agonists, 4 pyrethroids, 3 metals, 2 cholinesterase inhibitors, 2 nicotinic acetylcholine receptor agonists, valproic acid, verapamil, and fluoxetine. Seven compounds expected to have no effect on neuronal function were tested as "negatives" (glyphosate, acetaminophen, salicylic acid, paraquat, saccharin, d-sorbitol and amoxicillin). Following collection of 33 min of baseline activity, chemical effects (50 µM or highest soluble concentration) were recorded for 33 min. Twenty of the positives altered the mean network spike rate by more than the 14% threshold (two standard deviations from the mean for DMSO control). The three positives without effect were bifenthrin, nicotine and imidacloprid. None of the negative compounds caused a change in activity beyond the threshold. Based on these results, the mwMEA assay has both high sensitivity (87% identification of positive compounds) and specificity (100% identification of negative compounds). These experiments demonstrate the capacity of mwMEAs to screen compounds for neurotoxic effects mediated by a broad variety of mechanisms.

Computational estimation of errors generated by lumping of physiologically-based pharmacokinetic (PBPK) interaction models of inhaled complex chemical mixtures.

Many cases of environmental contamination result in concurrent or sequential exposure to more than one chemical. However, limitations of available resources make it unlikely that experimental toxicology will provide health risk information about all the possible mixtures to which humans or other species may be exposed. As such, utilizing computational models in order to make toxicological predictions is a useful tool in complementing experimental efforts which examine mixtures in health risk assessment. This paper outlines a novel mathematical method which reduces the complexity of a mixtures model and increases computational efficiency via a biologically-based lumping methodology (BBLM). In contrast to previous chemical lumping methodologies, BBLM allows the computation of error as a measure of the difference between the lumped simulation based on BBLM and the full mathematical model. As a consequence, the modeler has the opportunity to find the optimal configuration in the tradeoff between simplification and accuracy in order to determine an acceptable number and composition of lumped chemicals. To demonstrate this method, lumped equations based on a typical inhalation physiologically-based pharmacokinetic (PBPK) model assuming a competitive inhibition interaction mechanism are developed for a mixture of arbitrary size. The novel methodology is further tested using literature data for a mixture of 10 volatile organic chemicals (VOCs). Through simulation of these chemicals, BBLM is shown to produce good approximations when compared to the unlumped simulation and experimental data.